#!/usr/bin/perl -w

use DBI; 

# Create a Markov Chain for a text file and store it in a mysql database.
# Usage: cat file | genmarkov dbtable db_pass_for_root [nodel]
# nodel will prevent the program from doing DELETE FROM dbtable
# You must have a mysql server on localhost with a database named markov.
# The table you are using must exist and should have a structure similar to the
# following:
#	CREATE TABLE test (
#	  prev varchar(255) DEFAULT '' NOT NULL,
#	  new varchar(255) DEFAULT '' NOT NULL,
#	  prob varchar(255) DEFAULT '' NOT NULL
#	);


$max_words = 8;

die "max_words too big!\n"	if $max_words > 9;

warn "Getting text...\n";

while (<STDIN>) { # Store the text in $text since we're going to be making lots
		  # of passes through it. ($max_words passes to be precise)
	$text .= $_;
}

warn "Text stored.\n";
warn "Getting statistics...\n";

$aword = '[^ \t\n]+';
$notaword = '[ \t\n]+';
$text =~ m/^([^ \t\n]+)/;
$firstword = $1;
$sequences{""}->{$firstword} = 1;
$count{""} = 1;
for ($i = 1; $i <= $max_words; $i++) {
	$regex = "(";
	$regex .= ($aword . $notaword) x ($i - 1);
	$regex .= "$aword)";
	$regex .= '$';
	$text =~ m/$regex/;
	$sequences{$1}->{$firstword} = 1;
	$count{$1} = 1;
}

for ($curr_words = 1; $curr_words <= $max_words; $curr_words++) {
	warn "Getting information for sequences of $curr_words words...\n";
	$prev = "";
	for ($offset = 0; $offset <= $curr_words; $offset++) {
		warn "Offset is $offset.\n";
		$thetext = $text;
		$regex = "^";
		$regex .= ($aword . $notaword) x $offset;
		$thetext =~ s/$regex//;
		$regex = "($aword";
		$regex .= ($notaword . $aword) x ($curr_words - 1);
		$regex .= ")$notaword($aword)";
		warn "Using regex: $regex\n";
		FOO: while ($thetext =~ m/$regex/g) {
			$prev = $1; $new = $2; 
			$prev =~ s/[\t\n ]+/ /g;
			$new =~ s/[\t\n ]+/ /g;
			next FOO if $prev eq "";
			$sequences{$prev}->{$new}++;
			$count{$prev}++;
			warn "Found $prev->$new.  It's value is now " . $sequences{$prev}->{$new} . ".\n";
		}
	}
}

warn "Calculating probabilities and storing results in db markob, table $ARGV[0]...\n";
# open WORDS, ">$ARGV[0].mkv" or die "Could not open $ARGV[0].mkv for output: $!\n";
# open PROBS, ">$ARGV[0].mkp" or die "Could not open $ARGV[0].mkp for output: $!\n";
# open DATA, ">$ARGV[0].mkd" or die "Could not open $ARGV[0].mkd for output: $!\n";
$dbh ||= DBI->connect("DBI:mysql:markov", "root", "fatasshoL3");
die "Unable to connect to SQL Server" unless $dbh;


foreach $prev (keys %sequences) {
	foreach $new (keys %{$sequences{$prev}}) {
		$theprev = $prev;
		$prob = $sequences{$theprev}->{$new} / $count{$theprev};
		warn "Probability of $new is $prob.\n";
		#print WORDS "$word\n";
		#print PROBS "$probabilities{$word}\n"; 
		$theprev =~ s/\\/\\\\/g; $theprev =~ s/\"/\\\"/g; 
		$new =~ s/\\/\\\\/g; $new =~ s/\"/\\\"/g; 
		my $sql = "INSERT INTO $ARGV[0] (prev, new, prob) VALUES (\"$theprev\", \"$new\", \"$prob\")\n";
		$dbh ||= DBI->connect("DBI:markov:$ARGV[0]", "root", "fatasshoL3");
		$dbh->do("DELETE FROM $ARGV[0]") unless !$ARGV[2];
		if(!$dbh->do($sql)) {
			print "ERROR: " . $dbh->errstr . " while attempting to do $sql.\n";
		}
	}
}
#close WORDS;
#close PROBS;
#close DATA;

$dbh->disconnect();
warn "All done!\n";